Path: utzoo!attcan!uunet!ncrlnk!ncrcae!hubcap!gatech!rutgers!tut.cis.ohio-state.edu!ucbvax!decwrl!shelby!polya!geddis
From: geddis@polya.Stanford.EDU (Donald F. Geddis)
Newsgroups: comp.ai
Subject: Re: Question on Chinese Room Argument
Message-ID: <7621@polya.Stanford.EDU>
Date: 11 Mar 89 05:48:45 GMT
References: <2233@tank.uchicago.edu>
Sender: Donald F. Geddis <geddis@polya.Stanford.EDU>
Reply-To: geddis@polya.Stanford.EDU (Donald F. Geddis)
Organization: Stanford University
Lines: 26

In article <2233@tank.uchicago.edu> staff_bob@gsbacd.uchicago.edu writes:
>What can possibly be gained from this debate over Searle's thought experiment?
>Can't we let that debate wait until
>we're a little bit closer to something called machine intelligence?
>					R.Kohout

Not quite.  If the Turing Test model of understanding, thinking, and
intelligence is accurate, then we must apply behavioral criteria to a system
to tell whether it does cognition.  And in that case, some current systems
are uncomfortably close (for humanist critics) to already thinking.

Eliza does a reasonable job at a very small (and cleverly chosen...) domain.
More realistically, some large expert systems give as good answers to typical
questions in their domains as human experts do.

Of course you could always say that "real" understanding requires ability in
all areas of human existence, not just some narrow field.  But then my
mother doesn't understand theoretical physics.  It's a spectrum of
possibilities, and we're all somewhere on there.  AI systems tend to clutter
the low end, but there's no sharp dividing line where version (n) isn't
intelligent, but version (n+1) is.

	-- Don
-- 
Geddis@Polya.Stanford.Edu
"We don't need no education.  We don't need no thought control." - Pink Floyd